Political ads once relied on grainy footage and earnest voiceovers. Those days have vanished. Now campaigns unleash streams of computer-generated imagery that blur the line between persuasion and fabrication. And nowhere has this shift shown up more starkly than in conservative circles, where candidates and their backers have embraced generative AI to mock opponents, distort records, and chase votes with minimal effort.
Take the recent Missouri county executive race in St. Charles. Conservative hopeful Jason Law blanketed local airwaves with a video that looked homemade yet carried the telltale gloss of synthetic creation. In it, an AI-rendered version of state Sen. Bill Eigel rubbed his palms together in cartoonish glee while a generic text-to-speech voice declared, “Steve, wake up, I’ve been telling people for 12 years that I’ll cut their personal property taxes, and they still believe me.” The target, Steve Ehlmann, appeared grinning as his AI counterpart admitted, “Property taxes have risen by 300 percent over the past 20 years and I just tell people there’s nothing I can do about it!” Law himself showed up as a stiff approximation, promising a “fresh start.” Local station First Alert 4 captured the candidate’s own reaction. “It’s frankly an insult to older voters who don’t know that AI exists.”
But Law didn’t invent the tactic. He drew inspiration from Spencer Pratt, the former reality television personality who ran for Los Angeles mayor and deployed similar AI-generated attack spots. Reports from ABC7 highlighted how Pratt’s efforts turned heads for their crude effectiveness. The pattern repeats. Low barriers to entry mean almost anyone with a laptop can produce this material. The results often land with voters before anyone notices the artifice.
Farther east in Kentucky the stakes rose higher. Independent political action committees funded AI-generated spots aimed at Rep. Thomas Massie, the libertarian-leaning Republican incumbent. One clip featured what appeared to be surveillance camera footage of Massie checking into a hotel alongside progressive Democrats Alexandria Ocasio-Cortez and Ilhan Omar. The tagline blared, “Thomas Massie caught in a throuple!” Massie didn’t mince words. He told Louisville Public Media, “It’s always the losing campaign that does the crazy crap.” He added that such tactics represent “frankly an insult to older voters who don’t know that AI exists.” Despite the backlash, or perhaps because of the attention it generated, the Trump-backed challenger Ed Gallrein secured 54 percent of the vote according to WLKY. Observers wondered aloud whether the synthetic smears tipped the scales.
These examples form part of a broader wave. The NPR examined how AI deepfakes and memes polluted the 2024 election cycle. Elon Musk shared a fake advertisement in which an AI clone of Kamala Harris’s voice described herself as “the ultimate diversity hire.” He posted it without noting its origins as parody. The clip spread rapidly. Musk and other Trump supporters also circulated AI-generated images portraying Harris in Soviet attire or showing Black Americans rallying behind Donald Trump. Trump himself amplified a cartoonish AI picture claiming Taylor Swift endorsed him. None of it required studio time or professional editors. Just prompts and a few clicks.
Researchers at the Alan Turing Institute tracked similar patterns in their November 2024 report. The analysis, available via CETaS, found that AI-enhanced disinformation in the U.S. presidential contest largely reinforced existing beliefs. Supporters of a candidate often amplified content that aligned with their views, even when it originated as synthetic media. Polarization acted as fuel. When prominent figures like Trump employed these methods, their base followed suit. The report noted that influential accounts boosting such material tended to belong to outspoken backers rather than neutral observers.
Concerns extend past amusement. The Brennan Center for Justice warned in a 2023 analysis that widely accessible AI tools could accelerate disinformation and create fresh hazards for democracy. Their piece, hosted at brennancenter.org, pointed to an early AI-generated video that showed President Biden announcing a national draft to support Ukraine. The clip initially carried a disclaimer but later circulated without it, racking up millions of views. Similar fabrications targeted Sen. Elizabeth Warren, depicting her calling for Republicans to lose voting rights. As tools improve, the risk grows that bad actors could suppress turnout or bypass existing election safeguards.
Graphika offered fresh perspective in March 2026. Its study of three recent elections demonstrated how state-linked operators now use generative AI to create high volumes of bilingual content tailored to local contexts. The Graphika report described networks that produced fabricated videos, automated text in native languages, and coordinated posting across fake accounts to simulate grassroots support. In one Latin American case, the same bot network pushed both conservative and progressive messages, sowing confusion rather than clear advocacy. The volume alone overwhelmed fact-checkers.
Legislators have begun to respond. In Georgia, lawmakers adopted a resolution in June 2026 urging the state to confront AI deepfakes in elections. Rep. Josh Gottheimer, a Democrat from New Jersey, highlighted risks from biased chatbots that pull from unreliable sources when answering voter questions. His June 2026 social media post called for federal agencies to coordinate protections ahead of upcoming midterms. Yet enforcement lags. Detection tools struggle with newer diffusion models that power today’s most convincing fakes. Early methods that looked for absent eye blinks in videos have become obsolete.
Campaign operatives defend the practice. They argue that AI simply levels the playing field against well-funded traditional media buys. Supporters on the right point out that left-leaning groups have dabbled in synthetic content too. The University of Chicago Harris School white paper from 2023, still relevant today, cautioned that both sides harbor suspicions of bias in these systems. Conservatives worry the tools tilt against them. Progressives fear the opposite. The document, available through Harris School resources, stressed that generative systems can degrade the information environment by creating deepfakes nearly indistinguishable from reality.
Public reaction splits along familiar lines. Some voters dismiss the clips as harmless satire. Others feel manipulated. Older demographics appear especially vulnerable. They lack familiarity with the technology and often encounter these ads on television or Facebook rather than platforms that flag synthetic media. One X user captured the frustration in June 2026, noting how obvious bots continue to flood feeds with AI-generated election commentary while real accounts face suspension. Another post from the same period described an AI-generated ad attacking a Texas Senate candidate by having a synthetic voice sing a parody song about “trans kids.” The sponsoring group, Citizens for Sanity, spent six figures on the buy. Critics countered that voters might reject the campaigns behind such tactics more than the candidates they targeted.
International examples add context. A 2025 CIVICUS report on deepfakes in elections described experiments in the Netherlands where participants viewed manipulated videos of a conservative politician delivering a fabricated anti-immigration rant. The content included false claims about crime rates tied to immigrants. Participants rated the deepfake no more believable than simpler edits, yet the cumulative effect of repeated exposure remains unknown. The study, hosted by CIVICUS, warned that political actors increasingly use online platforms to discredit journalists who question the authenticity of such material.
So what comes next. Platforms experiment with watermarks and disclosure requirements, but adoption stays uneven. Regulators debate mandatory labeling for political ads that use generative tools. Tech companies tout detection algorithms, yet the pace of innovation in creation outstrips defense. One academic paper published in 2025 through PMC examined policy options for combating AI-driven disinformation. It emphasized the need for coordinated responses across governments and platforms. The authors avoided optimistic forecasts. Instead they mapped incremental steps that might limit harm without stifling expression.
Campaigns show no sign of retreat. The cost of producing professional-looking video has dropped from thousands of dollars to pennies. Speed has increased. A candidate can test multiple versions of an attack ad in hours and iterate based on early engagement metrics. This agility favors those willing to push boundaries. Conservatives have proven particularly adept at harnessing the tools for cultural wedge issues. Memes featuring historical figures or celebrities in absurd scenarios generate shares that dwarf traditional messaging. The line between entertainment and electioneering fades.
Experts caution against panic. Not every synthetic clip sways outcomes. Many voters approach political content with built-in skepticism. Alignment with preexisting beliefs matters far more than production quality. The CETaS researchers found that disinformation succeeds most when it confirms what audiences already suspect. In that sense AI acts as an amplifier rather than a creator of conviction. But amplification at this scale carries consequences. When thousands of variations flood the information space, the signal disappears beneath the noise.
Thomas Massie captured one dimension of the problem. His primary fight revealed how even established politicians can fall victim to tactics once reserved for fringe operators. The “throuple” ad didn’t need to convince viewers of literal truth. It needed only to plant doubt and spark conversation. In the attention economy, that suffices. Gallrein’s victory suggested the strategy carried weight. Similar tests will multiply as more races unfold.
Jason Law’s Missouri effort offered a simpler case. The ad didn’t invent scandals. It exaggerated policy disagreements through caricature. Eigel and Ehlmann became punchlines delivered by digital puppets. Law framed it as clever contrast. Detractors called it lazy deception. Both descriptions contain truth. The technology rewards brevity and emotional punch over nuance. Complex policy discussions don’t survive the transformation into 15-second clips voiced by synthetic narrators.
Looking ahead, the 2026 midterms will serve as another proving ground. Georgia’s resolution signals growing state-level awareness. Federal lawmakers like Gottheimer push for proactive measures. Yet the tools evolve faster than rules can form. Open-source models proliferate. Anyone can fine-tune them on specific politicians’ voices and mannerisms. The barrier isn’t technical anymore. It’s ethical and perhaps legal, though statutes have yet to catch up.
One fragment stands out from the X conversations. “Pretty convinced that voters will be more turned off to the campaign using generative AI than to the actual candidates.” Time will test that hypothesis. Early evidence from Kentucky points the other direction. The insult to older voters that Massie described may matter less than the sheer volume of exposure. Familiarity breeds acceptance, even for artifice.
Political operatives on both sides watch closely. Those who master the medium earliest gain advantage. Those who hesitate risk irrelevance. The result is an arms race that prioritizes speed and virality over accuracy. Generative systems don’t just create ads. They reshape the incentives that govern modern campaigning. Whether voters ultimately punish or reward the practitioners will shape the next chapter of American elections.
WebProNews is an iEntry Publication